KitGuru KitGuru.net – Tech News | Hardware News | Hardware Reviews | IOS | Mobile | Gaming | Graphics Cards
Additional technical assistance: Peter McFarland and Jeremy Price.
Correctly testing power supplies is a complex procedure and KitGuru have configured a test bench which can deliver up to a 2,000 watt DC load. Due to public requests we have changed our temperature settings recently – previously we rated with ambient temperatures at 25C, we have increased ambient temperatures by 10c (to 35c) in our environment to greater reflect warmer internal chassis conditions.
We use combinations of the following hardware:
• SunMoon SM-268
• CSI3710A Programmable DC load (+3.3V and +5V outputs)
• CSI3711A Programmable DC load (+12V1, +12V2, +12V3, and +12V4)
• Extech Power Analyzer
• Extech MultiMaster MM570 digital multimeter
• SkyTronic DSL 2 Digital Sound Level Meter (6-130dBa)
• Digital oscilloscope (20M S/s with 12 Bit ADC)
• Variable Autotransformer, 1.4 KVA
We are combining 12V output into a single result.
|
DC Output Load Regulation
|
||||||||||
|
Combined DC Load |
+3.3V
|
+5V
|
+12V
|
+5VSB
|
-12V | |||||
|
A
|
V
|
A
|
V
|
A
|
V
|
A
|
V
|
A | V | |
|
152W
|
2.07
|
3.40
|
2.04
|
5.02
|
10.11
|
12.18
|
0.50
|
5.01
|
0.20
|
-12.05
|
|
270W
|
3.03
|
3.38
|
3.06
|
5.01
|
19.11
|
12.12
|
0.50
|
5.00
|
0.30
|
-12.01
|
|
400W
|
4.01
|
3.30
|
5.04
|
4.97
|
29.17
|
12.08
|
1.00
|
4.98
|
0.30
|
-12.00
|
| 523W |
6.04
|
3.27
|
7.04
|
4.93
|
38.16
|
12.00
|
1.50
|
4.96 |
0.30
|
-11.97
|
|
652W
|
8.07
|
3.24
|
9.01
|
4.87
|
48.18
|
11.88
|
2.50
|
4.95
|
0.30
|
-11.94
|
Output load regulation is solid holding all primary rails within 2.5% of the nominal voltage.
| OCZ ZT 650W | Maximum Load |
| 724.1W |
The power supply would shut down at 724.1W, gracefully. We found the protection circuitry to work very well.
Next we want to try Cross Loading. This basically means loads which are not balanced. If a PC for instance needs 500W on the +12V outputs but something like 30W via the combined 3.3V and +5V outputs then the voltage regulation can fluctuate badly.
| Cross Load Testing | +3.3V | +5V | +12V | -12V | +5VSB | |||||
| A | V | A | V | A | V | A | V | A | V | |
| 590W | 1.0 | 3.34 | 1.0 | 5.00 | 48.0 | 11.85 | 0.2 | -12.05 | 0.50 | 5.02 |
| 145W | 12.0 | 3.23 | 15.0 | 4.95 | 2.0 | 12.08 | 0.2 | -12.03 | 0.50 | 5.01 |
The supply held well across the output tests, with the +12V rail dipping to 11.85V when loaded with 48A quickly.
We then used an oscilloscope to measure AC ripple and noise present on the DC outputs. We set the oscilloscope time base to check for AC ripple at both high and low ends of the spectrum. ATX12V V2.2 specification for DC output ripple and noise is defined in the ATX 12V power supply design guide.
|
ATX12V Ver 2.2 Noise/Ripple Tolerance
|
|
|
Output
|
Ripple (mV p-p)
|
|
+3.3V
|
50
|
|
+5V
|
50
|
|
+12V1
|
120
|
|
+12V2
|
120
|
|
-12V
|
120
|
|
+5VSB
|
50
|
Obviously when measuring AC noise and ripple on the DC outputs the cleaner (less recorded) means we have a better end result. We measured this AC signal amplitude to see how closely the unit complied with the ATX standard.
| AC Ripple (mV p-p) | ||||
| DC Load | +3.3V | +5V | +12V | 5VSB |
| 152W | 5 | 5 | 10 | 5 |
| 270W | 5 | 10 | 15 | 10 |
| 400W | 10 | 10 | 15 | 15 |
| 523W | 10 | 15 | 25 | 20 |
| 652W | 10 | 20 | 35 | 25 |
Ripple results all fall within the tolerance guidelines above, and they are impressive all round. The 12V+ output peaks at 35mV under full load.
|
Efficiency (%)
|
|
|
152W
|
85.42
|
|
270W
|
87.33
|
|
400W
|
88.45
|
|
523W
|
86.49
|
|
652W
|
85.65
|
These results are very strong for an 80 Plus Bronze rated power supply, peaking at over 88 percent efficiency at around 50 percent load. At full load the efficiency drops to 85.65 percent.
We take the issue of noise very seriously at KitGuru and this is why we have built a special home brew system as a reference point when we test noise levels of various components. Why do this? Well this means we can eliminate secondary noise pollution in the test room and concentrate on components we are testing. It also brings us slightly closer to industry standards, such as DIN 45635.
Today to test the Power Supply we have taken it into our acoustics room environment and have set our SkyTronic DSL 2 Digital Sound Level Meter (6-130dBa) one meter away from the unit. We have no other fans running so we can effectively measure just the noise from the unit itself.
As this can be a little confusing for people, here are various dBa ratings in with real world situations to help describe the various levels.
KitGuru noise guide
10dBA – Normal Breathing/Rustling Leaves
20-25dBA – Whisper
30dBA – High Quality Computer fan
40dBA – A Bubbling Brook, or a Refridgerator
50dBA – Normal Conversation
60dBA – Laughter
70dBA – Vacuum Cleaner or Hairdryer
80dBA – City Traffic or a Garbage Disposal
90dBA – Motorcycle or Lawnmower
100dBA – MP3 Player at maximum output
110dBA – Orchestra
120dBA – Front row rock concert/Jet Engine
130dBA – Threshold of Pain
140dBA – Military Jet takeoff/Gunshot (close range)
160dBA – Instant Perforation of eardrum
|
Noise (dBA)
|
|
|
152W
|
29.9
|
|
270W
|
30.3
|
|
400W
|
33.5
|
|
523W
|
34.3
|
| 652W | 36.5 |
The ZT 650W is relatively quiet under around 400W load when the fan spins up a little, becoming audible. When we get into the last 20% of ultimate power delivery it is spinning quite fast and is clearly audible. It wouldn't be classed as annoying, but for better acoustics performance, we would prefer to keep the power supply around the 500W output range (or lower) under load. This is more practical anyway because if you need 600W of constant power output you should be aiming higher in the OCZ range.
|
Temperature (c)
|
||
|
Intake
|
Exhaust
|
|
|
152W
|
35
|
37
|
|
270W
|
36
|
38
|
|
400W
|
39
|
43
|
|
523W
|
43
|
49
|
|
652W
|
46
|
53
|
The large 140mm has no problem maintaining good internal ambient temperatures. Going on these results we feel OCZ could have relaxed the fan profile a little to lower the noise emissions.
|
Maximum load
|
Efficiency
|
|
724.1W
|
83.3
|
Pushing the PSU above its rated limits generates an ultimate efficiency level of around 83.3%. This is not a viable ‘real world’ situation, but its interesting nonetheless.
